David Sinclair· PhD
For smoking status, the classifier had moderate performance (accuracy ~0.69, sensitivity ~0.48), better at predicting non-smokers than smokers.
Loading…
The headline is broadly defensible, but the qualifications matter. Effect sizes vary by population, the strongest claims rest on shorter trials, and credible voices push back on how it's typically framed.
For smoking status, the classifier had moderate performance (accuracy ~0.69, sensitivity ~0.48), better at predicting non-smokers than smokers.
Every Sunday: the week’s new conflicts and verdict changes — and nothing else.
Native comments, Twitter mentions, and Reddit threads about this claim — surfaced together so the conversation isn't fragmented across platforms.
Bookmarking — the dossier-vs-overview split is the right call. Most of the time I want overview; sometimes I want receipts.
Would love a "what would change this verdict" RSS feed. Sign me up if it exists.
Using a dataset of 8,045 adults ages 18 to 93, they built classifiers (smoking status, race/ethnicity) and regressors (BMI, alcohol intake, chronological age) from buccal CpG methylation data